The term fast Fourier transform (FFT) refers to an efficient
implementation of the discrete Fourier transform (DFT) for highly
compositeA.1 transform
lengths . When computing the DFT as a set of
inner products of
length
each, the computational complexity is
. When
is an integer power of 2, a Cooley-Tukey FFT algorithm delivers
complexity
, where
denotes the log-base-2 of
.
Such FFT algorithms were evidently first used by Gauss in 1805
[29] and rediscovered in the 1960s by Cooley and Tukey
[14].
Pointers to FFT software are given in §A.7.A.2